Trial lecture
Linguistic representations and their role in the era of LLMs
Time and place: January 30, 2026 11:15 AM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus.
Main research findings
In an era of information overload, transforming unstructured news into structured knowledge is a fundamental challenge for natural language processing. Event extraction, the process of identifying specific occurrences and the people or organizations involved, is essential for tracking global developments and detecting emerging crises. By automating these processes, we can significantly improve the efficiency of news analysis and the consistency of information tracking across diverse media landscapes.
This thesis advances both the methodologies and resources available for event extraction in the news domain. The research shifts from treating events as isolated labels toward representing them as unified semantic graphs that capture the full context of how entities and events interact. To address the gap in support for smaller languages, the work introduces new annotated datasets and investigates generative AI techniques that excel in settings where training data is limited.
The findings demonstrate that these sophisticated models produce more detailed and accurate representations of complex news narratives than traditional approaches. Additionally, the thesis shows how this structured analysis can be applied to evaluate and improve the factual consistency of automated news summaries. Ultimately, this research offers a robust framework for building more reliable and linguistically inclusive systems for global news monitoring.
Adjudication committee
- Professor Antske Fokkens, Vrije Universiteit Amsterdam, The Netherlands
- Professor Marco Kuhlmann, Linköping University, Sweden
- Professor Martin Giese, Department of Informatics, University of Oslo, Norway
Supervisors
Professor Lilja Øvrelid, University of Oslo, Norway
Professor Erik Velldal, University of Oslo, Norway
Associate Professor Samia Touileb, University of Bergen, Norway
Chair of defence
- Associate Professor Ellen Munthe-Kaas, Department of Informatics, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat